Calhoun: The NPS Institutional Archive

DSpace Repository

Reports and Technical Reports All Technical Reports Collection

1976-03

A multiattribute approach to measure quality

of health care

Giauque, William C.

Monterey, California. Naval Postgraduate School

http://hdl.handle.net/10945/30000

Downloaded from NPS Archive: Calhoun

wical repoht section

S4AVAL POSTGRADUATE SCHOOLgAWfO«NIA 939*0

NPS55Gi76031

NAVAL POSTGRADUATE SCHOOL

Monterey, California

A MULTIATTRIBUTE APPROACH TO MEASURE

OF HEALTH CARE

by

Wil 1 iam C. Giauque

March 1976

QUALITY

FEDDOCSD 208.14/2:

NPS-55GI76031

Approved for public release; distribution unlimited

spared for:

ice of Naval Researchington, Virginia 22217

NAVAL POSTGRADUATE SCHOOLMonterey, California

Rear Admiral Isham Linder Jack R. BorstingSuperintendent Provost

The work reported herein was supported in part by the Foundation ResearchOffice of Naval Research.

Reproduction of all or part of this report is authorized.

This report was prepared by:

UNCLASSIFIEDSECURITY CLASSIFICATION OF THIS PAGE (Whan Data Entered)

REPORT DOCUMENTATION PAGE READ INSTRUCTIONSBEFORE COMPLETING FORM

1 REPORT NUMBER

NPS55Gi 76031

2. GOVT ACCESSION NO. 3. RECIPIENT'S CATALOG NUMBER

4. TITLE (and Subtitle)

A Mul tiattribute Approach to Measure Qualityof Health Care

5. TYPE OF REPORT a PERIOO COVERED

Technical Report6. PERFORMING ORG. REPORT NUMBER

7. AUTHORf*;

Will iam C. Giauque

8. CONTRACT OR GRANT NUMBERfa;

9. PERFORMING ORGANIZATION NAME AND ADDRESS

Naval Postgraduate SchoolMonterey, California 93940

10. PROGRAM ELEMENT. PROJECT, TASKAREA ft WORK UNIT NUMBERS

61152N,RR 000-01010N0001476WR60052

11. CONTROLLING OFFICE NAME AND ADDRESS

Office of Naval Research

12. REPORT DATE

March 197613. NUMBER OF PAGES

14. MONITORING AGENCY NAME ft ADDRESSC" different from Controlling Office) 15. SECURITY CLASS, (of this report)

Unclassified15a. DEC LASSIFI CATION/ DOWN GRADING

SCHEDULE

16. DISTRIBUTION ST ATEMENT (of this Report)

Approved for public release; distribution unlimited.

17. DISTRIBUTION STATEMENT (of the abstract entered In Block 20, 11 different from Report)

16. SUPPLEMENTARY NOTES

19. KEY WORDS (Continue on reverse aide If necessary and Identify by block number)

Medical Care Decision Analysis

Quality of Care Multivariate Utility

20. ABSTRACT (Continue on reverse side It necessary and Identify by block number)

The problem of measuring the quality of health care is one of themost evasive, yet important, problems in medical administration. Outcomemeasures, the ultimate validators of care, suffer from a number of draw-backs, including uncertainty and the multiplicity of possible outcomecriteria. Measures of the medical decision process are easier to administer,but suffer from the lack of agreement on optimal processes for many

(Over)

dd ,;FORM 1.-79AN 73 "473 EDITION OF 1 NOV 65 IS OBSOLETE

S/N 0102-014- 6601|

UNCLASSIFIED

SECURITY CLASSIFICATION OF THIS PAGE (Whan Data Entered)

UNCLASSIFIED-U.CUW1TY CLASSIFICATION OF THIS PAGEfWhon Datm Entered)

20. situations. Measures of the physical or administrative structure ofan organization are easiest from a data collection standpoint, but

the optimal structure for quality care is not always clear. Multi-attribute utility (MAU) analysis can potentially resolve many of

the issues involved in quality assurance, including the onesspecifically mentioned above. In this paper the potential contri-butions of MAU analysis are outlined, a number of MAU studiescontributing to quality measurement discussed, and suggestions forquality assurance systems made.

SECURITY CLASSIFICATION OF THIS PAGEfWhen Dmtm Entered)

ABSTRACT

X'ne picbiem cf measuring the ^ualix.^ of health care is one

cl the iiiott tvasivc, yet important, problems in aeUical

aaministra t j on . Cutcome mesGureii, measures of the medical

decision process, and measures 01 the physical or

dumi:iatiative structure of an organization, the three major

approaches to quality measurement, all suffer from various

didw^ack^. Multia ttribute utility (MAU) analysis car.

potentially resolve many or the issues involved in quality

assurance. In this paper the potential contributions cf riAd

analysis are outlined, a number or MAU studies con tr irutinc

to quality measurement discussed, and suggestions for

quality assurance systems made.

a HUL'xIAITRIBOlE UIIiLITY API ROACH re KEj»SUfe£ ^JALIi* Or

KEALTri CAP;?

Willi am C. Giaugue D. B. a.

Assistant ProfessoL

Naval Postgraduate School

A LluLTIAITRIBu'lii UTILITY APPEOAvJri TO EEASUeE jUAtlTY Or

HEALTH CASE

Ine need to develop a useful measure ot tat quality of

medical cart has become iucree.sini.jly important in r^c^nt

years. A number or reasons ror this need can be cited,

including the necessity of determining t:ie quality of. care

delivered t)v innovative medical care stems, tne increased

interest m and use of paramedical professionals in

delivering care, tae need to control medical coits without

sacrificing quaJity, the increasing role of third parties in

paying tne costs of medical care, and the neeu to certify

institutions engaged in medical care delivery. fne,problem

or hew to treasure quality of care, however, is an elusive

one. Quality measurement requires one to cciii^ar^ multiple

etiectiveness criteria, to do so consistently, and to do so

ia an uncertain, complex en viornment . These considerations

suggest the potential userulness of muitiattribute utility

(lIAU) analysis in deriving quality standards. Eefore

discussing specifically the manner in which M'AU analysis

cculd ccntritute co this area, a orief summary of the

capacities cf MAC analysis is given, then soxe cf tne

problems involved in defining and measuring quality cf care

are discussed. The potential role of .MAU analysis ir.

attacking seme of these issues is indicated. Finally, a

suggested quality measurement system is outlined.

1. Summary of MAU Analysis

The tern MhU Analysis, as used in this t aper, refers to

the normative system of decision analysis as expounded in a

uumoer of sources, among them Eairfa 1, North*, Jchlaifer 3

,

and Brpwn et al* , together with t.ne results of

muitiattribute utility theory as discussed by Keen *y 5 , ° ,7

,8

,

r isnburn 9fl°

fl *, Fisnburn and Keenly 12

, and Faryubar,* 3,

auicng others. In essence, multiattribute utility theoretic

results allow one, given certain assump tion.s concerning

one's utility structure, to develop a ad express a utility

function ovei multiple attributes. Ihis results in the

development or a single measure of "goodness" which

suit Diarizes multiple, possibly conflicting, measures.

Furtnur, as ricst indicated by Von Neumann ^nd

rtcrgenstera* 4, the expected value of such a utility function

can ue useo as a decision criterion under uncertainty.

Decision analytic techniques then indicate how even complex

decision problems can be structured and analyzed, using a

utility function to guide the analysis. Final results can

include nqt only optimal decision strategies, but such data

as the value of additional information and the sensitivity

or results tc analytical inputs.

II. Defining and Measuring quality of Medical Care

with this preliminary summary of the analytic tools

completed, let us turn to the problem at hand, measuring the

guality of medical care. Defining "quality of medical care"

is itself a task cf remarkable difficulty. Donabedian 15

points out that in practice guaiity can be almost anything

anyone wishes it tc r>e, thus any discussion of the subject

is potentially plagued with misunderstanding. "Quality of

cars" is clearly not a unitary concept but a composite of

many, sometimes conflicting, desidtrata which must be

simultaneously considered. There is agreement on some of

the factors that go into guality success of the care in

preserving or restoring health, efficiency in the use of

resources, prevention and alleviation cf physical and

psychological suiiering but there is wide disagreement

on how inclusive. such a list of factors should be and how

much emphasis is due each of the aspects. We might examine

this issue by considering first measures of quality oi care

for ao individual, then discussing the issues involved in

extending these concepts to groups oc societies.

Quality Measurement tor Individuals

For an individual the delivery of r.:ecicai care can oe

thougnt cf as afrecting, in a probabilistic sens€, the

health, t sycho logical, ana financial outcomes experienced bv

the person. A number of these outcomes are 4u1riT.ifj.abie,

sucli as morbidity, mortality, amount of money expended on

nealth-r ela t ed items, freedom fro.ii pain and discomfort, even

degree oi freedom from psychological stress. For

individuals then, the ultimate validators of quality exist

in more or less accessible form, although in embarrassing

richness. tvhen trying to use outcomes to measure guality

one must resolve a number of issues. Fir^>c, which outcome

measures are to be used, and hew are they to be combined?

In treating patients with certain chronic disabilities, for

example, treatment strategies can depend strongly on whether

one considers morbitity or mortality of primary importance.

Second, even if one succeeds in devising a satisfactory

unitary measure, the uncertainty and complexity of medical

processes make it difficult to determine an optimal strategy

for delivering care. Third, even if an optimal strategy

were available, the diagnostic skills, treatment skills,

patient mdnagement sKilis, even the mechanical skiIIs (e.g.

now quickly and easily can a hypodermic needle be inserted)

of the medical practitioners inv/olved in the delivery system

can have a great impact on the outcomes. Thus, a standard

cf comparison is required against which actual outcon.es can

be measured. Ideally the standard would be quantitative,

would control icr the preexisting condition cf the patient,

and would alio*- :cr the inherent uncertainty i:\ any medical

intervention. A final proolem with any quality control

system which defends on outcome measures is the time delay

often required for some outcomes to become manifest. The

succtba qf some tctatitnts is not fully known until years

have passed.

SAU analysis ctfecs the capacity of dealing with some

or these issues. The issut of determining which outcome

measures are to be used and the way in wnich they are to be

ccmcined can be resolvea by assessing a utility iunction

over those outcome measures the person considers relevant.

The practicality of this approach has been demonstrated in a

number of studies. Giauque and Peebles 16, in analyzing

stieptoccccal sore throat and rheumatic fever, assessed a

utility function ever ten measures, including cost factors,

the numter of days ill with streptococcal infection,

severity of antibiotic reactions if any, and the existence

and severity of acute rheumatic fever and chronic rheumatic

heart disease. Kiischer 17, in analyzing patient management

decisions fcr clezt palate, assessed utility functions ever

such "ncnguantifiable" factors as the degree of speecn and

hearing iffpeuiment and the degree or disfigurement remaining

after treatment, as well as the cost cf the treatment.

Kdj-emicK 18 assessed a utility function over costs, various

degrees of illness and inconvenience, reduction in

longevity, and tne possibility of deatn in analyzing

hypertension. Ginsberg and Offensend 1 * assess utilities in

analyzing a particular case of nack pain, aitnough th-^y

determine utilities directly for a limited number of

outcomes rather than specifically assessing a multiattribute

utility function. This approach is somewhat simpler tnan

the multiattribute approach, nut is limited in that only a

siall number cf specific outcomes can oe considered.

A potential proDlem which is not totally resolved by

'A Ad techniques is that of whose utility runction should

guide the treatment of a patient. Clearly the patient

nimself should be the primary choice, rmt there are

situations where a patient's preferences may need to be

saioiainattd to an overaxl societal need, for exa» ri€ in

imposinq a quarantine. This issue may be or more

theoretical than practical importance if ail persons

involved navt utility structures implying identical courses

of action. In the stidies by Giaugue and Peebles 16 and

Krischer 17 tnis was indeed the case. Giaugue and Peebles

ce^otted that utilities assessed ccoin patients, doctors,

nurse practitioners, and public health officials raxie<3 froa

individual tc individual but not in any systematic way from

group tc gictc, and that in any case the solution was so

robust as to give identical optimal strategies for each

assessor. Krischer reported that the utilities assessed by

ail respendants to a questionnaire were very close tc each

other. However one cannot always count on results being

tnis tortuitous. In case of conflicting strategies, one

solution would ue to choose that which maximizes the group

utility of the entire society. In the case of quarantine,

for example, the disutility of the quarantine fgx the

individual irust be compared to the utility of disease

avcidance by the remaining population. This general lSbiie

is di-DCUosed furthur in the next section z>± tnis paper.

The prctlem of determining optimal strategies for

delivering care can dlso be attacked through MaU Analysis,

tacn or the studies cited above was decision oriented, in

that the optinal strategies for administering diagnostic and

treatment procedures were determined. A number of

additional cecisicn analytic studies of medical problems

could also tie cited. Ginsberg 20 analyzed the

pleural- effusion syndrome, expounding in addition a general

analytic framework for medical analysis. Lusted* 1,

Jacguez 22, Gcrry £3 ,? 4

, ana Gorry and Barnett^ 5 discuss

diagnostic approaches utilizing concepts of decision theory,

while betdgne and Goiry 26, Schwartz et ai* 7

, and GoLcy et

s l 28 discuss additional concepts in medical decision

analysis. Krischer 17 (Section 1.2) contains a useful

summary of nany of these papers. Giaugue 29 contains a

summary of decision analytic studies in non-medical, as well

as medical areas. Thus the feasibility of using decision

analytic techniques to structure and resolve coirplex,

uncertain problems has been demonstrated, but there remains

the issue ci practicality, determining whether an analysis

is worth the 1 not i ncohseguen tial time ar d trouble it taxes

to carry it cut. Often the analysis wculd net be worthwhile

for a single'' individual, but may be justifitd if the results

could be applied to entire groups of patients. Kost ci the

studies cited above were indeed intended to appl> tc mest or

all patients falling within certain classes. In some cf the

studies an attempt. was made to identiry those patient

characteristics which would affect the derived optimal

strategy. Giaugue and Peebles 16 for example examined the

effects cf patient age, days since onset of the symptomatic

streptococcal infection, and prior history of penicillin

reactions. Kapernick 18 controlled for patient age and

general patient health. Such studies can be considered

preliminary attempts to establish decision standards which

are controlled fcr the preexisting condition of the patient,

but clearly a good cieal more needs to be done tefore

definitive standards can be said to exist.

The third issue, estaolishing outcome standards to

cootiol for practitioner skill, cannot be done on an

individual patient basis due to the stochastic nature cf the

medical process. Just as good decisions do not guarantee

good outcomes, so good procedures administered with the

utmost skill, even when combined witn good decisions, cannot

guarantee good outcomes. The MA'li Analytic technigues

discussed abeve do, nowevt=r, establish average occurrence

rates for various outcomes. These data could potentially be

used as a fasis for a control system, but this would have to

bt over wary tdtienti, rather than for an inaii/Ldual ca

la addition, the optimal strategies can themselves serve as

standards, tut at iTocesj stanaarus rather tuan ouicon^

standards. Process measurement oilers a number of

advantages ever outcome measurement. first, results of

process measurement are available relatively soon,

immediately following the care delivery if necessary.

Second, iiiocess measurement attempts to directly assess the

quality or the decisions made, thus allowing one a standard

which dees not involve uncertainty. ihe uncertainty

regarding outcomes is automatically accounted for in setting

the standard. finally, many process measurements are

concrete, either in terms of whether or noi . a r t l c u a r

service was ferf orated in an individual case, or in terrcs of

statistical measures, such as the proportion of the

population reached, the volume of services rendered, and the

costs or service. Fiany suggested 4uality control techniques

are ouiit arcund process measures. Foist 30 suggests using

^recess stancards constructed by MAU analysis in deter nining

settlements in malpractice suits. Flagle 51 suggests nine

measures cf process of care, covering the areas of

inclusiveness, adequacy of content, and proa uc t i vit y

.

Dcnabedian 3 ^ contains an extensive discussion of the issues

involved in irocess measurement, many or which relate to the

practicality cf a measurement system for large groups. This

leads us into a consideration of issues involved in quality

measurement fcr groups and societies.

Quality Measurement for Grou^i.

Sup^cse we have successfully assessed utilities from

and determined optimal treatment strategies for each meaner

01 a particular grcup, and we are now faced with decisions

which may affect all momriers or tue grcup. Decisions

concerning care can still be made for eacn meaner of the

grcup individually, but these individual decisions are

constrained cy, and in some cases guided oy, group

decisions. Ihe design or the delivery system, lor example*

can limit access to even Dasic medical care cy some ^arrs of

the population. A societal decision to extend tae

availability ci udsic care througn say a ^ysteii. or low cost

neighborhood clinics mignt result in better care fcr son?

people, hut one might legitimately ask whether this would"

result in higher overall quality than building, ror example,

a dialysis urit fcr those with Kidney disease. In either

case the failure of the group, to provide certain rescurces

may effectively limit the options open to the individuals.

We are new faced wit ft tne problem of determining a so-called

social welfare function, a measure wnich summarizes the

welfare, or utility, or the group as a whole. If sucn a

function could r»e found then our choice of group action

could be guided by it; but there are some difficult

theoretic ar.d practical proolems involved. Kirkwood 33 Chap.

II contains a summary of these issues. Briefly, it is

possible tc define a. social welfare function, given the

utilities fcr each memoer of the group, hut only it some

restrictions are wet. Perhaps the most convenient form is

that given by Harsanyi 34, which gives the group utility of

any alternative as the weighted sum of the utilities of each

individual fcr the alternative. Required conditions are

that both the group utility and the individual utilities

obey the ven Neumann - Morgenstern axioms of cardinal

utility cr their eguxvalent, and that if twe situations are

indifferent frcm the standpoint of all individuals, then

tney are indifferent for the group as a wnole. Kirkwood 33

generalizes these results by applying the concepts of

pairwise preferential independence and mutual utility

independence, as discussed in Keeney 5,6

,7

,8

, tc group

utility structures. The more general formulations developed

by Kirkwcoi also construct the group utility function by

multiplyirg tne individual utilities uy weighting constants,

then adding and/cr multiplying the weighted utilities.

Assuming these lor mulations give reasonable approximations

in leal conditions, one must only assign weights tc = acn

individual, in efrect determining whose preferences should

count the nest, to get the group utility luncnon. Jiving

each person in the group an eguai ^eignt is one crvious

possibility, althougn this raises scrae interesting g.uestions

(should mere weight be given to those uno give the greatest

financial support to the system, snouid age or general

health affect weights, how should preferences or persons *:;o

deliver the care be accounted for, etc.)*

biven the possibilities or tnis methodology in

determining optimal strategies for the group, tne practical

problem regains 01 how to set up a ^uaiity control system

whicn exploits them. As in tne case of individual guaiity

control the outcomes experienced bj tne group can DC jsea to

mtasure the overall effectiveness of a system, witn the

signiricant aavantage that uncertainty , in a large sample,

can re at least partially accounted for. Average occurrence

rates, determined oy the optimal strategies cnosen, can

ssrve as gtality standards. ricwever, the other major

difficult)' with outcome measures, the time delay often

occurring uetween treatment and the final observation c!: all

outcomes, still remains. £ven if the delay is not excessive

there are formidable problems in gatnering aata on all

pertinent outcomes, especially once the patient 1 saves tne

site. we can, though, use process standards to determine

the quality at least of the decision making, though perhaps

not the sKill ol the practitioners in performing tne

processes

.

Donaredian 32 , in tus extensive discussion of process

standardo, points cut that it we attempt to define standards

tor every pcssinle situation, even allowing the possitiiity

of setting meaningful standards, we would become hopelessly

bogged down in endless detail. Clearly we cannot hope to

predefine optimal actions ror each possible problem for each

possible patient. However, i»e can determine, tor those

medical problems which are important, where uncertainty

exists concerning which among significantly different

courses ci action is best, the relative desirabilities of

different strategies, and determina the sensitivity of the

choice of strategy to patient cnaracteristics . fnus,

critical factors are identified, allowing the practitioner

to rocus his attention and use his judgment on relative wail

defined issues. In addition, retrospective analysis can ne

used for guality control after the fact, either in medical

audits or in lawsuits, as suggested by Forst 30.

lac mc: c han-tca 1 a-id ddcinistrati/e problems of deoignj.ng

an ongoing, systematic guality assessment program i.^.u.'1.

formidable. Supposing that outcome and/or process standards

exist fcr at least some areas, how are the data measuring

actual outcomes and processes to be collected? Overall

statistics cr some outcome measures, such as mortality, are

sometimes available, cut data en other measures may be

completely lacking. Patient records are generally sketcay,

incomplete, and difficult to access. Recollections of

patients and practitioners are subject to Dias,

inaccuracies, and incompleteness, while medical

practitioners, particularly physicians, are loth to te too

critical of colleagues. The mechanisms for conducting

process reviews also lead to problems. Case reviews are

expensive and suffer from the lack of good source data.

Direct ofcservation of a practitioner's activities is aiso

expensive and is apt to change the practitioner's beaavior.

In addition, the observer may not know as much as the

practitioner in sone areas, particularly concerning patient

histories, thus possibly leading to inaccurate judgaents.

Statistical indices, are easy to review, but may be

difficult tc collect, to identify with a particular system

10

or. Cai-O, and to control for patient characteristic; . r:ir

these reasons, stuuctaral measures usual!} supplement

uancoine measures ana process measures in quality assessiue it

.

otru^tuie ifiea.sur.es examine the physical, r jo less ioiid l

,

and operational structures of i -ist it'it ions in which aedicii

care taxes r iace. As defined oy bonabedian 1 5, structure

measures are ''concerned witn such things as the adequacy of

racilities and equipment, tae qualifications of medical

staff and their organization; tae administrative structure

and operations of programs and institutions providing care;

fiscal organisation and the like." Structure measures can

re relative easily and cheaply made, and the results ueavailable quickly. In addition, they are fairly concrete,

at least in part, thus are mor-= amenable tc control through

guidelines and legislation than otner measures. The lajor

drawback of structure measures is the rather tenuous

connection between structure and outcome. There are ^jme

instances wher^ a connection can be drawn; if certain

processes car be identified as desirable in treating a giv.-r.

disease and if those processes require certain types of

training cr equipment, then clearly the presence of that

equipment cr ci adequately trained personnel is a necessary,

though net sufficient, condition to -juoa quality care. Even

in tnis case, however, the actual delivery of high ^uality

care is not assured. In more general situations, there can

be significant differences cf opinion concerning the

ccntr inu tions cf various structural measures to tne ^ulaity

of care.

III. Sujcested Quality Control Systems

The setting in which medical care is being delivered

hat a great impact on tne feasible mechanisms for

adn mistering a guality control system. Jn this section we

will discuss three types or setting, a large cc11c.roi4.ed

group, a snail informal group or private practice, and

juasi-medical institutions. There are, 01 course, other

possioilities, but these three will suffice 10 illustrate

the major issues.

large Controlled Groups

A large controlled group is typified by many military

health care crgani zatxon,-.:,- as well as by some large civilian

group or public health practices. The major characteristic

of such groups is the existence of a recognized authority or

chain cf authority in administrative matters, and at least

to some extent in medical matters. Sucn organizations serve

sizable populations and visually, though not always,

individual relationships between tne patient and the health

pract it icr.er are not strong. This creates a need (not

always fulfilled) for a good patient record system, while

the existence of the central authority provides the means to

design and inplement such a system, Furtnur, the size of

such groups makes innovative methods of health care delivery

notn possible and important, and the need to evaluate the

resulting guality 01 care is especially acute. For such

settings, separate levels of quality control based on

outcome treasures, process measures, and structure measures

are suggested. First, outcome data can o-j collected through

the patient record system. To minimize the difficulties

caused by the time lag between treatment and the observation

cf cutccires, "indicator" presenting symptom complexes could

be selected and statistics on the outcomes fcr patients with

12

tnose presenting symptoms gathered. The indicator sy,.i.aoKi

ccapiexes should be such that outcomes couicl be observed

within a fairly short time, and should ue ccmmon enougn to

allow reasonable sample sizes for each practitioner. Trie

term "syaptcm complex", incidently, is used in place or

"disease" since trie patient presents a complex of syor r tomc

to tne uedicai system, aau determining tne cause ct tae

syaptoms, the disease, is part of the diagnostic problem.

Scire possible syaptcm complexes are 35 headache, lowei back

pain, cons tipa tio r , and obesity. For each symptom ccjiplex

chosen, MAU analysis could be used to determine which

diagnostic/treatment processes are optimal, to examine the

effects of patient characteristics upon optimal treatment

cncices and expected outcomes, to establish the expected

rreguency of various outcomes, and to determine wnat outcome

data snould le collected.

Tne second type of quality control appropriate in large

controlled institutions is process control. As discussed

earlier in this paper, MAU analysis can lead to process

standards for a number of symprom complexes. In an

institution of this sort process control can re i mpieoiented

taiougn reexamination of selected patients, by patient

interviews, and tnrough patient record audits. Again the

MAU analysis would indicate tne data that should be captured

en patient records in order to ma*e the audits complete.

Fiagie 31, in discussing process standards, suggests nine

criteria, nanely

Measures ci inciusiveness;

proportion of the population reached,

proportion of health problems covered,

Measures of content;

ccmplet eness of services,

rationality of services,

responsiveness of services,

numaneness of services,

13

Measures of productivity;

volume cf services rendered,

health productivity , and

cctts of service.

Tnt discussion acove directly addresses issues in tne areas

ct completeness and rationality of services, but the MAU

techniques cculd just as easily be used to determine, for

exaaple, appropriate levels of inclusi veness.

Tne third level of control, cased on structure

measures, can te applied oy ensuring tnat the capability for

good quality care exists, in that appropriate equipment for

the desired procedures is available and personnel are a^Le

tc carry cut the procedures. MAU analysis can both indicate

what the optimal procedures are, and quantify tne degree of

quality given up if less that optimal facilities are

available

.

Snail informal Group or Private Practice

Problems of quality control in this setting are

extremely difficult, as indicated by the literature (cf,

Donabeaian 15,3 ^, iflagle 31 ). Major issues are the lack of

objective, complete data on patient symptoms, treatments,

and outcomes, the difficulty of accessing what patient data

is availarle, the lack of any person or ^roup with the

recognized authority to make quality judgments concerning

private practices, and a long tradition against meddling

with a physician's "private" affairs. It would clearly be

very dirficuit tc gather outcome data ror such practices,

but one mignt reasonably hope to utilize some types of

process control. It seems reasonaole, for exaaple, to

contuct "patient audits" from time to tirot. A sample of

patients, -sitner chosen randomly or selected by specific

symptom complexes, could ce interviewed, and the

14

tes 1- treating nt sequence reconstructs . through pat

records and the recollections ot the patient and the

physician. Inis or course would require a change in

attitude cu the part of many physicians. Measures based dii

structure would fce much easier to construct and wouic 1.7

an indication of a t least ttie potential for good quality

care. Certification anu recertif icat ion examinations and

continuing education requirements can oe ana are ueing u

to insure technical competence, wmie cnecklists of required

laroratory facilities, medical instrumentation, examination

facilities, dnd office procedures can insure an adequate

physical and administrative enviornmen t.

Certification ot Quasi-Medical Institutions

There is currently a need for medical certification

prcctdureir fcr such quasi-medical institutions as nursing

hemes, rest homes, sanatorium^, and the like. In terms of

ccntrollaoilit y, such institutions fail between large groups

ana private practices. Ihe right or an authority, usually

the state, tc examine and question medical standards is

recognized, but the direct authority evident in, say, a

military medical raciiity is lacking. In these situations,

quality standards based on structure measures are certainly

appropriate and reasiole. If an institution requeots

certification to admit and treat a certain type of patient,

clearly the physical facilities and medical stafr required

for quality treatment would have to be available. The major

questions nere generally pertain to the degree ol capability

required. Is it necessary to have a bull time pnysician in

a nursing hcue, cr is it sufficient to have cne available on

call? Need regular physical examinations be provided? How

often? Is a nurse acle to deliver quality cai__j in this

setting, cr a physician's assistant, 01 is a physician

required? MA'O analysis offers the capauixity of exanining

15

any nunibct of such questions.

Process measures can be made wore effective in

inst it ut lcnal than private practices since tne state car.

reasonaoly impose standards of patient record keeping, and

can reserve the rignt to conduct patient audits from time 10

time. In institutions most patients are pnysicaiiy available

for extended periods, so it becomes feasioie to conduct

independent medical examinations. Finally the extended

nature of institutional care allows for the gathering of

seme outcome data. If cure rates, length of stay, or other

outcome measures for a particular institution are abnormal*

a mere intensive investigation, such as a patient audit,

could be conducted.

1b

iilBLIGiJBAPHY

Raiiia, H., decision Analysis, Addison- Lesley , Redding

[lass. 1968

4.

North, E.ri., "A Tutorial Introduction to Decision

rhecr/," i^£^ i£^i»sdct ions in System s^ij^nc^ milQXtzznctics, Vol. oSC-4, No. 3 (Sept. 19o3) pp. ^00-210

Schlaiier, a.O., Analysis 21. £<£C±sions ^fider

UB2iiI^iiiiIr McGraw-Hill, Hew York, 1968

brown, 5, V., Kahr, A.S., and Petersen, C. , Lecision

Anaiisis for the Manager, Holt, Rinenart and fcinston,

New i o i k , 19 7 4

5. Keeney, R.L., "Multidimensional Utility Functions:

inecry, Assessment and Applications, 1 ' Tecnhical ;ort

No. 43, Cjierations Research Center, KIT, Cambridge,

dass. Ccxouec 19o9

b. Keeney, B.L., "Utility Independence and Preferences tor

Multiattiibuted Consequences," Operations Research Vol.

19, pp. 216-261 (1972)

7. Keeney, R.L., "Utility Functions for Multiattr lbuttd

Consequences," Management Science 18: pp. 27b- 287, 1972

6. Keeney, R . L . , "Multiplicative Utility Functions,"

leenhical Report No. 70, operations Research Center,

Mil, Cambridge, tiass. March 197 2

9. fishnurn, t.C, Decision and Value Theory, Wiley, New

KorK 1964

17

10. Fishfcurn, P.C. Utility Theory for Iccisi_on ^^illSi*

Wiley, New Yoik, 1970

11. Fisfccutn, P.L., "von Neumann -Mor gen^tern Utility

Functions on Two Attributes," Operations Research Vol.

22 He. 1 (Jan. -Feb. 1974) pp. 35-45

12. Fishburn, P.C. arid Keeney, R.L. "Generalized Utility

Independence and Some Implications," Tecnnicai heport

No. 93, Operations Research Center, MIT, CambLidge,

Mass. April 1974

13. fciquhai, P.H., "Fractional riypercute Decompositions

oi Multiattritute Utility Functions," Technical Report

222, Department oi Operations Research, College of

Engineering, Cornell University, Ithica, Neu York,

August 1S74

14. von Neumann, J. and Morgenstern, 0., Theory_ of Ga^es

&H& iccncmic Eehavior, 3rd ed., nfiiey , 1953

15. Donabedian, A., "evaluating the Quality of Medical

Care," Milrjank Memorial Fund Ojiar terl.y_, Vol. 44, No. 3,

1966

16. Giaugue, rt.C. and Peebles, T.C., "Application of

Multidiaensional Utility Theory in Determining Optimal

Test-Treatment Strategies for Streptococcal Sore Ihroat

and iineunatic Fever," Technical Report NPS-55Gi74 10 1

,

Naval Ecstgraduate School, Monterey, Calif., October

1974

17. Krischer, J. P., "An Analysis or Patient Management

Decisior.s a~ Applied to Cleft Palate," Center of

Research in Computing Tecnnology, Harvard University,

18

Cambridge, Mass., Dec. 1974

18. Kapernick, H.3., "Medical Decision Analysis - 4n

Appiic dtioi, in Hypertension, " Masters Thesis, i^vai

Postgraduate Scnool, Monterey, Calif., March 1975

19. Ginsberg, A.S., and Offensend, F.I., "An Application

or Decision Theory to a Medical Diagnostic-Tr eatni

Problem, " IJJ2 Transact ions in System ^citncj ^n^

^iternetics, Vol. 3SC-4, No. 3 (Sept. 196£) pp. 355-_>b2

20. Ginsberg, A.S., Decision Analysis in Clinical Patient

flanageiaent With an Amplication to the Pl^ur^I-Eif asion

iyndrcme, Rand, H-751-3C/NLM, 1971

21. Lusted, I.B., introdactioa to Medical Decision ^aKinj,

Charles C. TLcmas, Illinois, 196a

22. Jacguez, J. A., The Diagnostic ^I^cess, Ann Arbor,

Michigan, Malloy Lituographmg, Inc., 1964

23 Gorii G.A., "Strategies for Computer Aided

Diagnosis," Math^ Biosciences 2:293-316, 19od

24. Gorry, G.A., "Modelling the Diagnostic Process,"

Journal cf ^odical Eaucation, 45;z93-302, 1970

2d. Golly, G.A. and Sarnett, G.O., "Ixpenence with a

Model cf Sequential Diagnosis," Computers and

xiicfiiedical Research, 2:490-507, 19o8

2b Betagae, N.L., and Gorry, G.A., "na ten a ting Judgmental

Decision Making for a Serious

Management Science 17:342 1-434, 1971

Medical Problem,"

27 Schwartz, W.B. et al, "Decision Analysis and Clinical

19

Judgment, " American Journal of M§^i£ine r :>5; 45 9-472,

Oct. 19 1 3

28. Goiry, et al , "Decision Analysis as tne Basis for

Computer-Aided Management of Acute Renal Failure,"

American Journal of Medicine, 55:473-4 84, October 197 3

29. Giau^ue, w.C, "Organizational Decision MaKiiij,"

Technical Keport NPS-55Gi75091 , Naval Postgraduate

School, Monterey, Calif., September 1975

30. forst, B.cl., "Decision Analysis and Medical

Malpractice," Chelations Research Vol. 22 So. 1

(Jan. -Feb. 1974) pp. 1 - 12

31. Fiagle, CD., "Assessment of Quality cf tne Process of

Care," in Ccllin, ii.F. (sd.) Proceedings or an

International Conference in Health Technology Systems,

ORSh Health Applications Section, 1974

32. Donabedian, A., "Promoting Quality through Evaluating

the Erccess cf Patient Care," Medical Care, Vol. 6, No.

3, May-June, 1968

33. Kirkwccd, C.W., "Decision Analysis Incorporating

Preferences of Groups," Technical Report No. 74,

Operaticns Research Center, MIT, Cambridge, Mass. June

1972

34. Harsanyi, J.C., "Cardinal Welfare, Individualistic

Ethics, and Interpersonal Comparisons of Utility," J^

ZQl± Iccncmy, Vol. 63, pp. 309-21 (1955)

35. for tnis paragraph I gratefully acknowledge the ideas

of Cr. Jcsepn D. Bloom, Dr. Ralph H. Palumbo, and Dl.

James J. " Quinn, all Capt., MC, USW, of the Naval

20

Regional Kedical Center, San Diego, Caii.

21

INITIAL DISTRIBUTION LIS!

No. of Copies

Defense Cccunent at icn CenterC direr on Stat icnAlexandria, Virginia 22314

12

Dean cf SesearchCode 023Naval Postgraduate ScnoolMcnterey, California 93940

Litrary (Code 0212)Naval Postgraduate SchoolMcnterey, California 93940

Lilrary (Code 55)Naval Postgraduate SchoolMcnterey, California 93940

Professor R . K. ButterworthPrcfessor C. E. DerrProfessor C. K. EoyangPicfessoi D. P. GdverProfessor w. C. GiaugueProfessor P. J. ParkerProfessor C. A. PetersonProressor E. A. SchradyPrcfessor M. G. SovereignProfessor M. U. ThomasPrcfessor D- R. WhippleDepartment cf Operations Research

and Administrative SciencesNaval Postgraduate SchoolMonterey, California 93940

1

1

1

1

101

1

1

1

1

1

E. David ArramsAssociate Director fcr ResearchHealth Planning Research Services,550 Finetcwn ReadFort Washington, PA. 19034

Inc

Carla AutelliLitraryCentro DiagncsticoVia Saint Eon 2020147 MilancITALY

I taliano

Prcfessor L. I. BaronGraduate Schcol or ManagementDepartment of Managerial Economic:Northwestern UniversityEvanston, Illinois b020'J

and Decision Sciences

Professor CscarDepartamentcM .-.i T, ,, r- ,- -! A t, A

deBa rroslndustrias1. ' 1 -Universidad de Chile

Casiila 2777 - SantiagoCHILE

22

Di . John JDepartmentCollege cfUniversityNotre Dame

Eernar docf ManagementBusiness Administrationcf Notre DaraeIn d 1 a n a 4655b

Cart. Joseph D. Blcom, MC, USNChairirar, Department or InternalNaval Regional Medical CenterSan Jieqc, California 92134

Medic in*

J c h n DAatui .:

Schooi of Pharmac,Ajturn, Alabama 3563

Be waanUniversity

Dr. J . mi . B u s nDirector. Division of Health Policy ana ManagementMuir College, Bidg . 2-D, it id. 2346Department oi Community MedicineSchool ci MedicineUniversity cf California, San DiegoLa Joila, California 92037

Z a n e K . CampfellDirector, Ccrj-crateOmark Industries2100 S. E. Milport RoadPortland, Oregon 97222

Planning

Prcfessor Jared L. CononThe Department of Geography andTne Johns Hotkins UniversityBaltimore, Maryland- 2 121

8

1

Environmental Engineering

Prcfessor G. Col sonEcclt D* Admir istra tionFacuite de DroitUniversity de liegePlace du AX Aout, 324GC0 liege3EIGIUM

des Affaires

Consad Research Corporation1^1 u. Highland AvenuePittsburgh, Pennsylvania 15206Attn: Mrs. Sheila Steger, Librarian

Proressor Rolf A. BeiningerThe University of MichiganSchool or Putlic HealthDepartment or EnvironmentalAnn Arnor, Michigan 43104

and Industrial health

Dr. Edward N. DodscnGeneral Research CorporationScience and Technology Division5363 Holiister AvenueP. 0. Box 35B7Santa Barrara, California 93105

2 A

Picfessor Janes S. DyerUniversity of California, Los AngelesGiaduate Schc.cl of ManagementLos Angeles, California 90024

Dr. Charles Eby17001 rt'ocdaife DriveSilver Spring, MD 20904

Professor Lynn a. ivansBaylor College of MedicineTexas Medical CenterHouston. Texas 770 25

Dr. Ptie: h. JarguharInfer is-* tic n Sciences DeptIhe iiANC Corporation1700 Main StreetSanta Monica, CA 90406

Professor Chaiies FlagleDepartment of CpeLations ResearchThe Johns Hopkins UniversityBaltimore, Maryland 21205

Dr. Brian E. EcrstSenior Research Analystinstitute for law and Social Research1125 15th St. N. W.Suite 6 2

L

WashingtcE, C. C. 20005

Dr. Charles C. GibsonBxecntivt DirectorTulsa Area Healtn andRe cm 6 073 2*4 iidin MallTulsa, Oklancua 74 10 3

Hospital Planning Council

Pccfessor John K. GohaganWashington UniversityProgram in Technology and Human AftairsBox 1106Attn. Gohaga n/WilliamsSaint Louis, Missouri 63130

Dr. A. J. GoldmanChief, Operations ResearchDivision 205.02Administrat icr. Building, RooruU.S. Department of CommerceNational Bureau of StandardsWasnington, C. C. 20234

A4 2 8

Dr. William A. Goiomski59 Bast Van Euren StreetC.iicago, Illinois 60605

Dr. Paul GrayUniversity cl Southern CaliforniaSchcol of Business AdministrationDepartment of Quantitative Business Analysis

24

University ParkLos Angeles, Calixcrnia 90007

Prctesscr Douglas D. GregoryOakland UniversitySchcoi ot Economics and Manag<Rochester, Michigan 48063

iment

Professor' David H. GustaisonDepartment of Industrial EngineeringUniversity of Wise en sinMadison, Wisconsin 5370o

Dr. Herman EDirector, StatisticalMedical ResearchSan don, Inc.East Hancver, N.

HamotServices

Department

J. 07936

Ll . Edward L. Hannanhealth Studies CenterBailey HallUnicn CcilegeSchenectady, N. ¥". 12308

Dr. Michael C. HarrisStair EconcuistSouthwest Michigan Comprehen,Partnership fcr~ Health517 Ship St. Rm. 7Peoples State BanK Blag.St. Joseph, Mich. 49085

1

;ive Health Planning Association

Professor Brian D. HenshailDecartment of Management StudiesIhs University of AucklandPrivate £agAuc klandNEK ZEALAND

Eartara C. Hicks, M. S. S. tf.

Research AssociateBo106Department of Eehavioral Sciences and vommunityThe University of Connecticut Health CenterFar mington, Ccnnecticut 06032

Health

Prcfessor G. Eavid HugnesGraduate Schcoi of Business AdministrationUniversity of North CarolinaCnapel Hill, Ncrtn Carolina 27514

Prcfessor Ralph KeeneyMass. Institute of TechnologyOperations Research Center, ft

Cambridge, Mass. 0213924-215

Professor Kerry E. KiipatricKDirectorJ. Hillis Miller Health CenterUniversity cf floridaGanesville, Florida 32610

25

Dr. L. F. 'Krys' KrystynakSenior Scientific AdvisorDepartment of Health, Education,Public Helath ServiceHealth Resoruces AdministrationRcckville, Maryland 20852

and Welfare

Professor Leonard R. LambersonIndustrial urgine-enng ana Operations ResearchCclio-je of EngineeringWayne state UniversityDetroit, Michigan 43202

Dr. Agnes D. LivingoodPublic Policy Analysis DirectorUrtan Studiss CenterGarlencourt CampusAlta Vista ReadUniversity cf LouisvilleLouisville, Kentucxy 40205

Professor Stuart H. MannThe Pennsylvania State UniversityDivision or Man-Environment RelationsS-126 Human Development BuildingUniversity Park, Pennsylvania 16802

Dr. Benedict F. MassellResearch EirectorHouse of the Good Samaritan25 Hinny StreetBoston, Mass. 02115

Dr. James E. MathesonDirector, Decision AnalysisStanford Research InstituteHenlc Park, CA 94025

Group

Dr. Doug McClureVA Hospital (6643350 La JcllaSan Diego, CA 92161

4/1 12)Village Drive

Pic lessor Richard MeyerHarvard Eusmess SchoolSoldiers Eield Road.Boston, Ma: 02163

Dr. Allen C. Miller, IIIStanford Eesearcn InstituteMenlc Park, CA 94025

Professor Holmes E. MillerSchcol of MaiagementRensselaer Pclytechnic InstituteTrey, New York 12131

Dr. Matthew P.. MirabileAssociate DirectorHealth Studies Center

26

Bailey HallUnion Coil eyeSchenectady, New York 1230 a

Ihe MITRE Cc l perat ionPO Fox 20fcBedford, Mass. 02173Attn. Eetty Erown

PiciessorSchool oiDepartmentUhi vei sitA u h e r s t

,

Joel N. Worsebusiness Administrationot General Business and

y oi MassachusettsMassachusetts 01002

finance

Frciessor K. i. . K. NairDepartment ci business Adaii;; isttaticnThe Universi ty of New BrunswickFrederictcn, N. S.CANAEA

NationalLyndon BHouston, Texas 7 7 o 5 8Attn; Edwin Eu^nick FM3

Aeicnautics ana Space AdministrationJohnson Space Center

Dr. Joan f. C'EanncnProject AdministratorHealth Services Research CenterKaiser Foundation Hospitals47G7 S. E. Hawthorne Blvd.Portland, cregon 97215

Dr. Greg1 J7 v'8 N.Redmond-,

Pa w j. ion^Oth Place

Washington 98195

Dr. Thomas C . PeeblesGeneral Medical Associates101 River *cadWeston, Mass. 02193

Professor DavidHealth ServicesGraduate Schcol

W. PenticoAdministrationof Public Health

Universitv of PittsburghIPittsburgh, PA 15zt>1

Prcfessot Carles E, PerezUniversity ci PittsburghDepartment ci Industrxai EngineeringSystems Management Engineering and Operations10*43 Benedum HallPittsburgh, Pa. 152o1

Research

Mr. James A. Pc^eGuiircrd College Urban Center5HC0 West Friendly Ave.Greensboro. North Carolina 27410

Professor Howard naiffa

27

Harvard Business SchoolSoldiers Field rtoadBoston, Mass. 02163

Dr. W. I. Rasmussen, Code 3400Naval .Ciectrcnics LaDcrctory CenterSan Diegc, California 92152

Professor Richard A. ReidThe University of New MexicoSchool of Business and AdministrativeAlt uguergue, New nexico 87131

Sciences

Professor Eduard J. RisingUniversity of MassachusettsAmherst, Mass. 01u02

Professor Stephen E. RobertsIndiana University School of MedicineDepartment of Communit.y Health SciencesMarion Count} General Hospital960 Locke StreetIndianapolis, Indiana 4o202

Ms. Judith A. RocheVector Research, Incorporated£.C. Box 15CfaAnn Arbor, Kichigar 48106

Ptcfessor Caii A. BothGraduate School or ManagementDepartment of Managerial EconomicsNorthwestern UniversityEvanston, Illinois 60201

1

ana Decision Sciences

Professor Hovard P. BuddCalifornia State College, Bakersfield9001 Stockdaie HighwayBdKersfield, Calircrnia 93309

Dr. Donald F. SchallerMedical DirectorArizona Health Plan925 E. McDowell BeadPncenix, Arizona 85006

I. M. SircyezhinChoi., Department of Economic CyberneticsInstitute cf Finance and Economy191023Leningrad, USSB

Professor Surrata K. SenThe University of RochesterThe Graduate school or ManagementRochester, New York 14627

Professor Keith V. SmithUniversity of California, Los anagele;Graduate school cf Managementi.os rinoej.es CA 90024

28

J . E. SpaldingAssociate PiciessorCclxeqe ct Business AdministrationNcrth'l'exas State UniversityDenton, Texas 76205

Professor Charles lilcuinDepartment cl Health Admini^t rat ionUniversity ct MontrealCas*i Postale 6128Wort real, Canada 101

Professor V. SrinivasanStar lord UniversityGraduate school or businessStanford, California 94305

Dr. Lawrence M . ThallExecutive Directorrr,ali and Earnydt Associates3178 Bluett BeadAnn Aroor, Michigan 46105

Inc

Picressor Bfriam TurbanDirector, Health Hanagement ProgramsSchool of business and Organizationalflorida International UniversityTaaiami TrailMiairi, Florida 33144

Sciences

L-rcfessor J. P. van GigchDepartment cf ManagementSchcoi of Business and Public AdministrationCalifornia State University, Sacramento60C0 J StreetSacramento, California 95819

Prciessor Wa^ne WinstonIndiana UniversityGraduate Schcoi or BusinessSchool cf Business Building81ccn!ingtcr , Indiana 47401

Ms. Sue -WclffBritish Gas CcrporationKe^earch anc Development326 High EclrcrnLondon, EisGLANE

Division

Picfessor Tuncei ri . YegulalpKruraL Schccl ct Mines916 S. V,. MuddColumbia Uriversitv5 20 W. 12Cth St.New York, N. Y. 10 027

Prcfesscr David Zalkind8m. 264 tept. of Health Administratioi

North Carolina 27514Univer si ty~ cf -North CarolinaChapel Hill,

29

Professor Milan ZelenyColumbia UniversityGraduate School of BusinessNew York, N.Y. 10027

Iioifc^aor a. 1. ZiembaFaculty of Ccmirerce and Business AdministrationUniversity of British Columbia2075 Westrock PlaceVancouver, B.C., Canada

Dr. Russell Ziiriuer manb-K Dynamics, Inc.15625 Shady Grove RoadRockville, Maryland 20850

30

U: '169

DUDLEY KNOX LIBRARY - RESEARCH REPORTS

5 6853 01071211 -HPS